IiV SITU
OBSERVATIONS OF THE FORMATION MARTENSITE IN STAINLESS STEEL
OF
J. W. BROOKS, M. H. LORETTO and R. E. SMALLMAN Deparrment of Physical Metallurgy and Science of Materials. University of Birmingham. P.O. Box 363, Birmingham Bl5 ITT (R~ceired 3 April 1979; in rmised form 18 June 1979) Abstract-Specimens of stainless steel have been deformed at room temperature. or cooled to below .if, in an HVEM and the formation of martensite observed. The *II, temperature. orientation relationships and habit planes of the martensites formed in specimens thicker than 0.5 pm were found to be identical to those of the bulk material. It has been shown that the e-martensite occurs in regions where appropriately, but usually irregularly. spaced stacking faults are formed, while x-martensite nucleation is associated with dislocation pile-ups on the active slip plane. avons d&form& des &hantiilons d’acier inoxydable B la temperature ambiante ou nous les avons refroidis au-dessous de &f, darts un MEHT, et nous avons observt la formation de la martensite. Nous avons trouve que la temperature AI,, les relations d’orientation et les plans d’accolement pour les martensites formkes dans des Cchantillons d’Cpaisseur suerieure B O,j pm Btaient identiques a ceux du matCriau massif. La martensite f se forme dans des r&ions oti l’on a des dkfauts d’empilement espaces irrtgulikement mais de manikre appropriie, alors que la germination de la martensite I est associke A des empilements de dislocations dans le plan de glissement actif. R&wm&-Nous
Zusammenfassung-Proben rostfreien Stahls wurden bei Raumtemperatur verformt oder im Hochspannungselektronenmikroskop unterhalb Af, abgehiihlt, urn die Martcnsitbildung zu beobachtcn. .LI,-Temperatur. Orientierungsbeziehungen und Habitebenen in Proben mit einer Dicke grii0er 0,s m waren identisch mit denjenigen im vollen Material. Es wird gezeigt. dai7 der E-IMartensit in Bereichen auftritt. in denen giinstig, aber iiblicherweise unregelmPBig verteilte Stapelfehler gebildet wurden. Der r-hlartensit hBngt mit Versetzungsaufstauunsen auf der aktiven Cleitebene zusammen.
1. INTRODUCTION The morphology and crystallography of martensitic phases have been extensively studied in a number of materials but the mechanisms of nucleation and growth are still unclear. This is assumed to be due to the small size of the martensite nucleus and to the very rapid kinetics of the transformation. The majority of electron microscopy of martensitic structures has been carried out on material transformed in the bulk prior to ob~rvation. However, a fuII understanding of the transformation mechanism cannot be reached using this technique due to the difficulty of differentiating between crystal defects formed before the transformation and those arising from the shape change. Thus the true sequence of events can be established only from in situ observations of the transformation. Unfortunately the thin foils suitable for conventional 100 kV microscopy are essentially two-dimensional and consequently it is likely that the results obtained would be of little relevance to the bulk transformation due to the small matrix constraint. Maki and Wayman [If carried out a thorough comparison of Fe-Ni and Fe-Ni-C martensites formed both in thin foils and in bulk using 100 kV electron microscopy and they concluded that the morphology and substructure of thin foil martensites
was rather difficult to character& due to the large differences in dislocation and twin distribution from one plate to another. They observed the same orientation relationship in both bulk and thin foil martensites; the austenite-martensite interface of the thin foil martensite was usually irregular and non-planar, whereas the bulk structures are character&d by smooth planar interfaces. Warlimont [ZJ, again using 100 kV microscopy, measured the effects of foil thickness on the iM, tem~rature and concluded that martensite nucleation in thin foiis is governed by two opposing effects. As the foil thickness decreases then MS decreases due to the fewer low energy nucleation sites available. However, this is offset by the decrease in energy required for nucleation and consequently MS increases for very thin foils. These results indicate that the behaviour of martensite in thin foils is more complex than that implied by the simple geometrical changes. Recent in siru experiments [3] on Fe-Ni-C martensites, carried out using HVEM, have shown that the structure of the martensite in the thicker parts of the foil was more representative of the bulk, as partially twinned plates formed in the lenticular shape characteristic of bulk Fe-Ni-C martensites. The orientation relationship and habit plane of the foil martensite were consistent with those of the bulk martensite of
1830 BROOKS. LOREJ--fO
AXD
SMALLMAN:
MARTENSITIC
FORiMATfON
IN STAINLESS STEEL
was produced by electrolytic polishing at IOV in a
,
13% perchloric acid and 87:4 glacial acetic acid solution at room temperature. The specimens were examined in an AEI EM7 high voltage electron microscope operating at 400 kV, in either a straining stage or a holder in which the specimen could be cooled to about - 150°C.
supporting arm.
Fig. 1. Geometry of in siru deformation specimens. The initial specimen thickness is 0.1 mm. The completed specimen is bonded onto a sliding CuiBe holder, using epoxy resin, prior to mounting on the microscope stage.
the same composition and the martensite formed in bursts without the subsequent isothermal growth observed in the thinner specimens examined by Maki and Wayman [l]. In the present investigation the formation of martensite in stainless steel was observed in an HVEM during both in situ cooling and in situ deformation experiments and the ~ystallo~aphy of the martensite and the morphoiogy of the lattice defects formed were compared with the bulk behaviour. These preliminary observations, which are reported in the present paper, showed that the .transformation behaviour in thick foils was identical to that observed in bulk specimens. On this basis a thorough investigation of the nucleation of both the z and E phases has been carried out which is reported in the following paper. 2. EXPERIMENTAL Three low carbon Fe-Cr-Ni steels (Table 1) prepared from high purity iron, ferrtihrome and ferronickel by induction melting under argon, were homogenised for 12 h at 12OO”C,also under argon. The alloys were then reduced by cold rolling with intermediate anneals at 1000°C in uacuo, to strips 0.25 mm thick. Some of the material was then partially transformed to martensite by cooling in liquid nitrogen. Specimens suitable for electron microscopy were prepared from 3 mm discs of transformed and um transformed material by electrolytic polishing at f4OV in a twin jet polishing cell using 3% perchloric acid, 35% butoxyethanol and 62% ethanol solution at -15°C. The austenitic specimens were subsequently transformed to martensite during in situ cooling experiments. The geometry of the specimens used for the in situ deformation experiments (see Fig. 1)
3.1. In situ and bulk defect strucf~res The microstructure of bulk-deformed low stacking fault energy (SFE) stainless steel is characterised by the presence of hexagonal (E) and body-centred cubic (z) martensite in a matrix of heavily faulted austenite (j). Similar defect structures are formed on cooling to the MS temperature as only partial transformation to martensite occurs and the retained austenite is heavily faulted [4]. The three alloys examined were all of low SFE which partially transformed to martensite when quenched into liquid nitrogen. The principal features of the retained austenite in these materials can be seen in Fig. 2, which shows the typical microstructure obtained on cooling in the bulk to the ‘ci, temperature. When in situ straining experiments were carried out on a fully annealed material with a low initial dislocation density, it was found that the majority of dislocations were produced at grain boundaries although some Frank-Read sources were observed. As deformation continued the dislocations propagated on (11 lf, planes and in many cases dissociated to form long extended stacking faults (Fig. 3). This gradually gave rise to a structure very similar to that formed during bulk deformation. The &fs temperature of 16f12 stainless steel has been reported to be -59°C [S] and during in situ cooling experiments very little happened to the defect structure until below this temperature. However, between -60 and -70°C the unit dislocations present in the material dissociated and many stacking faults formed from partial dislocations nucleated at grain boundaries (Fig. 4). These observations cannot be explained as thin foil effects as the majority of stacking faults were nucleated in the thicker regions of the specimen and not from the edges of the foil. It was thought that this behaviour might have been due to stresses caused by thermal contractions in the specimen holder. However, a stainless steel with a much lower Ms temperature was cooled down to - - 150°C and very little
Table 1. Alloy Composition in wt% Nominal Composition I6114 16/l? 1s;10
Cr
Ni
C
Si
S
P
16.20 14.00 co.02
Other elements 0.002; balance Fe.
Mn
MO
0.09 0.02 CO.06 0.02 co.01 0.02
Fig. 2. Typical defect axxtnre
formed in stainless steel on coo&q q is 200 and marker 0.5 pm.
in the bulk to MS. EC- jGl<].
f
Fig. 3. Dislocations and exrensive faulting prxtuced during B -_ [CN3]_ iY.31
in siru
deformarion of 18: 10 staia!sss s:ce!
i831! BROOKS, LORE-IT0
4. Eistended
AX= SlcfALLMW:
,MARTE%3-TlC FORMATION
K?i STAIXLESS hSTEEL
stacking faults formed at MS during in situ cooling of the 1612 alloy. By is ii 1 and marker is 0.5 pm.
dislocation activity occurred. The similarity between &fs and the temperature at which the faulted structures form indicates a close relationship with the martensite ~ansformation and in subsequent experiments it was observed that both x and E martensite nucleation occurred in association with these defects.
32.1. wrrartazsite. The transformation between the f.c.c. austenite and the c.p.h. c-martensite has been termed ‘fully coherent’ [6]. as the close-packed planes and directions in the two structures are paraliel. i.e. (.111):!(0001t [lIO]!.:[ lZlO& This orientation refationship Fas observed both in the ‘bulk’ e-martensite and in that produced during the in siru experiments. The hexagonal martensites formed during in sirirudeformation and coohng both nucleated from randomly spaced overlapping stacking faults and consequently the l-martensite =-as always
IPI II,
faulted in the early stages of the transformation. The e-martensite was usually formed from overlapping stacking faults, which nucleated at gram boundaries (Fig. 5), although some was produced during in situ deformation as a result of the interaction of dissociated slip dislocations with an obstacle, such as a grain boundary (Fig. 6). 3.22. ~-~ten~~te. The low carbon content of the three alioys ensured that the u-martensite formed was body-centred cubic. The orientation relationships were determined by taking composite diffraction patterns, using various known beam directions in the austenite. These were indexed consistently, since the sense of tilt between each beam direction couId be ascertained from the Kikuchi lines formed by the austenite matrix. Previous work on stainless steels indicated that the orientation relationship of the bulk martensite was usually close to the Kurdjumov-Sachs (K-S) relation. The standard variant of this is
Fig. 6. e-martensite nucieated, during in situ deformation. at a slip band-grain boundary intersection.
It was found that the diffraction data from both the bulk and in situ martensites could be indexed in accordance with the K-S relation and no orientation relationships peculiar to thin foils were observed. Figure 7 shows an r-martensite plate formed during in situ cooling which was indexed in terms of the standard variant of the K-S orientation relationship as [101], was approximate& parallel to [OTOJ, (Fig. 8) and subsequent tilting for -50” about [11 I], (parallel to [lOi],) towards [Oil], gave rise to the diffraction maxima corresponding to the [ii-& zone. in all the nucleation events observed the x-martensite formed at one of the activated slip planes and initiaby grew out on both sides, although the majority of subsequent growth took piace on one side onfy. An example of a partially transformed specimen, in which the trace of the ill if, plane associated with the nucleation event can be clearly seen is shown in Fig. 7 and it is interesting to note that, in terms of the standard variant of the K-S relation, this is the f I if&
plane which is parallel to the close-packed tlolt, plane in the r-marten&e. All the martensite plates analysed behaved in this manner which indicates that not only does the martensite in these materials nucleate from faulted defects but also illustrates that the close-packed planes in the two structures are parallel right from the start of the transformation. The habit plane could not be determined in the majority of eases as, by the time the martensite had become large enough to exhibit an identifiable planar interface, the matrix was so deformed that a crystaliographic analysis was not feasible. However, it was possible to establish the austenite-martensite interface of the r-martensite shown in Fig. 9 using trace analysis. The micrographs were indexed consistently with the K-S relation and the habit plane obtained was close to (2251,. The accuracy of this is at best 25’. Even so this agrees well with the habit plane found for martensite produced in the buik. Kelly [7] carried our a thorough investigation of the habit planes found
FiJz. ._ f
x -marr:xiw
formed during in siru cooling of the 16 12 allo?. The slip ban3 a~sociaizd lrith nxlcarion (A-B) is clearly visiblr. B _ nOI]. g is I I I and msrbcr is 0.i gm.
in Fe-Mn-Cr-Ni and Fe-Cr-Ni steels and observed a fairly- Iarge scatter around (1121, and (ES), which could not be attributed solely to ex~~enta1 error; some variation from {?%I, is not unreasonable. The crystallographic behaviour of the z and e-martensites formed during in situ experiments is, therefore. very similar to the bulk properties. This combined with the similar microstructures formed prior to the transformation indicates that the nucleation and growth processes observed are likely to be similar to those occurring during the macroscopic transformation.
3.3.1 E-n~~rrensirr. The close packed planes and directions in the tc.c. and c.p.h. structures are paralfei and consequently the transformation can be effected bq’ iaulting on alternate planes. Fujita and Ueda [a] pointed out that e-martensite could be formed from the regular overlapping of stacking faults produced by a specific mechanism or from random irregular overiapping processes. Yotasa [9], in a study of the f.c.c.4.c.p. transformation in cobalt, obsened that the number of stacking faults produced during cycling
around the ~ansformation temperature decreased with each cycie. This was interpreted as evidence for the existence of a pole mechanism which could not operate due ta the small foil thickness ( <3X0 .A,. However, since it was also found that unit dislocations did not dissociate on cooling it seems more likely that the decrease in the number ai extended stacking faults was due to surface effects pinning the existing dislocations. In the present work the f-martensite produced during in sirtt experiments aiways formed from randomly spaced overlapping fauits on the active slip plane and no evidence was found for a nucleation process involving a pole mechanism even in the thickest areas of the foii which could be rxamined in the HVEM tie. 12 pm with image intensification). When e-martensite was produced from stacking faults nucleated in regions of the foil which were too thick for electron penetration it was also faulted which again indicates that the faults were not formed by a pole type mechanism as this would give rise to regular faulting and hence relatively unfaultzd f-martmsite. Fujita and Ueda f8] observed simiiar behaviour in an IS 8 stainless stzel and suggested that ths transformation nucleates from irregularly spaced
Fig. S(a).
Fig. X(c).
Fig. 8. Diffraction patterns illustrating the orientation relationship between rhc atrstznirc and the xmarrensite shown in Fig. 7. (af + (b) 8’ off (Oil 1: and ! I i 1jr. Lc)[TOIt, ,,‘@iO),. faults which form at itis and then tirther faulting is induced on nearby {l 111, planes as this minimises the bulk free energy and the total energy of the stacking faults. A mechanism has been proposed [S] for producing relativefy unfaulted E-martensite which involved cross-sfip. However, cross-slip reactions, in general, are not a suitabte means of producing E-martensite due to the high energies of the dislocations
generated and this is supported by the experimental observations which show that e-martensite usually Forms on the operating slip ptane rather than the cross-slip plane. Thus +martensite nucleates from irregularly spaced bundles of stacking faults which gradually become perfect hexagonal b~stals as it is energeticalfy Favourable to generate !%nAtswhich give rise to the required ABAB stacking.
Fig. 9(a).
!ar&s can be formed at the austenite-martensite interface during growth and this would account for the highly dislocated substructure of the martensite. Ahhough this was the most common form of martensite embr)-o observed in this investigation it is suggested that nucfeation may presumably occur at any disfocation configuration which gives rise to a pseudo-marten&e b.c.c. structure. 4. Ct-3NCLUSfONS bee IO phone.
frc 111pu!rrt.
Fig. 10. h shear of I 6
(a) The nucleation of x and +martmsites has been observed during in situ straining and cooling some low SFE stainless steels. (b) The behaviour of the martensites produced in the thicker foils used in the HVEM is representative of the bulk material and consequently the nucleation events are thought to be typical. (c) Ali the nucleation events observed in the present work occurred in association with dislocation motion on {l ll), slip planes. It seems likely therefore that martensite always nucleates from suitable dislocation configurations which are generated by the stress associated with the free energy difference between the austenite and the martensite. Thus the kinetics of martensite nucleation can be explained without assuming the presence of pre-existing embryos as the required heterogeneities can be provided by the dislocations formed prior to the transformadon. Acknowlf~~ements-One of us would like to acknowledge an S.R.C. Studentship and support from the S.R.C. under B’RCi;‘58? is also gratefully acknowledged. REFERENCES I. T. Maki and C. M. Wayman. Acra metall. 25. 651 2. 3. 4. 5.
6. 7. 8. 9.
(1977). H. Warlimont. Metall. Trans. 2. 1847 (1971). T. N. Durlu, Scripta mrtall. 12, 603 (197gI. P. M. Kelly and J. Nutting, .!IISI 197. 199 (1961). J. F. Breedis, Tmns. AiME 230, 1583 (lw). J. W. Christian, The Theory of Transformation in Metals and Alloys. Pergamon Press, Oxford (1965). P. M. Kelly, .4cra 5ferull. 13, 63.5 (1965). H. Fujita and S. Ueda, dcca me&f. 20. 759 (1972). E. Votava, JIitl 90, 129 (1961).